By Parnian Taidi-Oracle on Jul 18, 2014
The value any ERP system can deliver depends on real-world performance, not just theoretical capabilities. Infor and Oracle recently conducted joint performance tests using Infor LN on Oracle SPARC T-series servers with customer-based transaction mixes, at the Oracle Solution Center in Linlithgow, UK. The measured results were 10X faster than previous records and showed near-linear scaling in both 2-tier and 3-tier deployments.
Infor LN is a uniquely powerful ERP solution for complex manufacturing, optimized for distributed global operations and equipped with the tools to improve operational performance in manufacturing plants, across the supply chain, and throughout the aftermarket service business. Oracle's SPARC T-servers running Oracle Solaris 11 are cost-effective, highly efficient cloud infrastructure platforms optimized to accelerate enterprise applications, delivering extreme performance and outstanding virtualization capabilities in a compact design.
Realistic OLTP test loads were generated using a wide range of transaction types to measure how well these combinations of hardware and software could perform while serving a large number of simultaneous users under different performance densities. These tests simulated larger user counts and transaction loads per core than most Infor LN customers require.
Two testing configurations, found in high-demand business environments, were used:
- 2-tier: The Oracle 11g R2 Database software and the Infor LN software were both deployed on the same physical server.
- 3-tier: The Oracle 11g R2 Database and the Infor LN software were deployed on separate physical servers.
The Infor LN application servers were connected via 40 Gbps InfiniBand links to the Oracle Database server which was connected to one Pillar Axiom 600 high performance, highly efficient modular storage system. All tests were conducted running Infor LN in unicode mode. When unicode mode is configured other character sets than standard ASCII - referred to as single byte mode - are supported. Because unicode characters need more resources to compute they influence application performance.
The results on the SPARC T5 server was measured to be 10X faster per processor than the previously best record performance measured for Infor LN on the M8000 16-processor (SPARC64 VII+) server. An Oracle M8000 16-processor server is approximately 10X more expensive than an Oracle SPARC T5-4 server. The combined price/performance benefits of running Infor LN on Oracle SPARC T5 servers prove to be extremely advantageous for customers.
Near Linear Scaling
Multiple configurations were tested and all showed near linear scaling for Infor LN:
• SPARC T5-4 for 2-tier deployments
• SPARC T4-4 for 2-tier deployments
• SPARC T4-4 for 3-tier deployments
2-tier Infor LN OLTP throughput comparison:
Infor LN Scaling:
These tests demonstrate that the combination of Infor LN and the Oracle SPARC T-Series hardware prove to be a highly reliable configuration offering, excellent performance with highly consistent scalability and broad flexibility for single-site or multi-site enterprise implementations for customers. In addition, the scalability of Infor LN on the SPARC platform can make it easier for an organization to accommodate more users or higher loads with additional servers or CPU cores.